Portail des publications scientifiques IMT Mines Alès
Not a member yet
5198 research outputs found
Sort by
Apprendre aux machines à comprendre l'eau : Adapter l'IA à la complexité des hydrosystèmes
No full textInternational audienc
Making the Best Use of High-Resolution Meteorological Forecasts for Flash Flood Forecasting Using ANN: Case Study on the Gardon de Mialet Basin
No full textInternational audienceIn Mediterranean regions, the difficulty of measuring and forecasting rainfall intensity, as well as the difficulty of identifying flood-generating processes, often lead to the use of statistical models, like neural networks. However, without the coupling with meteorological forecasts, current hydrological models are most often limited to a lead time equivalent to the response time of the basin, i.e. a few hours for small basins. The challenge is to increase this lead time, which is often too short for crisis management. As support for demonstration, a flood forecasting model for the Gardon de Mialet basin (Southern France) has been developed. The neural network model implemented (a Multilayer Perceptron) has been fed with the AROME high resolution model to produce forecasts with lead times of up to 24 hours, compared with 2 to 3 hours previously. The use of raw data predicted by AROME for the 47 rainfall events in the database produced flow results which were used as a basis for improvement. We therefore began by identifying the sources of forecast error (meteorological or hydrological models, or both models together), and then qualifying and quantifying meteorological forecast errors. At the end of this process, an approach is proposed for making an optimal use of weather forecasts from AROME, taking into account errors in synchronization, location and intensity of the rainfall
Multi-stage Multi-resolution Fusion for Accurate and Efficient Whole Slide Image Segmentation in Colorectal Cancer
No full textInternational audienceWhole slide image (WSI) segmentation plays a critical role in the precision medicine of colorectal cancer, as it enables detailed analysis of tumor morphology and microenvironment, which are essential for accurate diagnosis and treatment planning. However, this task faces two major challenges: (1) the gigapixel resolution of WSIs necessitates patch-based processing, leading to the loss of global contextual information; and (2) existing multi-resolution fusion methods suffer from the loss of positional and semantic information due to late-stage feature fusion, insufficient utilization of low-resolution contextual information, and high computational complexity, resulting in suboptimal segmentation performance. To address these issues, we propose a novel multi-stage multi-resolution fusion framework. First, we introduce a dual-branch encoder, where the frozen low-resolution branch efficiently captures global contextual information, while the trainable high-resolution branch preserves fine-grained spatial details. Second, we specifically design a decoder that first optimizes low-resolution features to mitigate the imbalance between dual-branch features, followed by stage-wise feature fusion to fully leverage the wide field of view of the low-resolution branch for enhanced segmentation results. Additionally, we incorporate a multi-scale feature fusion and optimization module to deeply refine features and improve the model’s segmentation performance. Our method demonstrates significant advantages in multi-class semantic segmentation tasks on a colorectal cancer WSI dataset. The code and model weights will be made publicly available to support clinical decision-making
Water–energy–food–ecosystem nexus: how to frame and how to govern
No full textInternational audienceThe food, energy, and water domains are strongly interrelated. The body of literature investigating these interlinkages developed into water–energy–food and, more recently, ecosystem (WEFE) nexus thinking. The WEFE nexus is concerned with cross-sectoral resource management requiring dedicated WEFE governance approaches. Among the existing WEFE nexus conceptualizations, the one that places the ecosystem at the center emphasizes the importance of ecosystem services for human well-being and as the basis for the functioning of the water, energy, and food domains. Such conceptualization, however, lacks clear definition and explanation of implications for WEFE nexus governance as well as practical tools for implementation. Accordingly, based on an in-depth analysis of WEFE nexus interlinkages, associated governance challenges, and practical experience, we propose the WEFE Nexus Governance Approach as an operationalization of the ecosystem-centric WEFE nexus conceptual framework. This approach takes policy coherence and stakeholder co-creation at its core, integrated with quantitative modeling of WEFE nexus interlinkages, and is organized in four steps: problem identification; formulation of substantive ambitions through stakeholder dialogue; embedding of ambitions and action into a stakeholder agreement; and implementation of actions. The approach is oriented to practical application and can be implemented by any actor (e.g., research institutes, governments, non-governmental organizations) having the ambition to initiate a stakeholders co-creation process toward nexus governance in a specific context. Furthermore, it can be applied to different scales and nexus domains depending on the identified nexus problems, while accounting for the vital role of ecosystem services. The approach’s applicability and needs for future research are discussed
Identification de symbioses éco-industrielles de la filière hydrogène : développements méthodologiques et application
No full textNational audienceCette étude, réalisée dans le cadre du projet Carnot Mines Hytrend, porte sur le développement et l’application d’une méthodologie d’identification et de caractérisation de synergies territoriales basées sur la coproduction d’hydrogène et d’oxygène par électrolyse. Les coûts de la technologie d’électrolyse sont élevés. Une façon d’atténuer les coûts est d’identifier des symbioses de l’hydrogène permettant la valorisation commerciale de l’oxygène comme coproduit, actuellement non pris en compte. S’il est relativement aisé d’identifier et de caractériser les synergies possibles pour de grandes zones ou de grands sites industriels, il est beaucoup plus difficile de le réaliser pour des petites installations décentralisées. Cette étude s’est donc plus précisément focalisée sur ce verrou
Hypotheses in Opportunistic Maintenance Modeling: A Critical and Systematic Literature Review
No full textInternational audienceBecause they account for realistic effects in opportunistic maintenance modeling, dependency hypotheses are extremely diverse in the literature. Despite recent reviews, a clear view of the dependency hypotheses is currently missing in the literature, especially regarding component interactions, resource constraints and human factors. In this paper, we provide a conceptual background on dependence modeling and the notion of maintenance opportunity. Then, a critical systematic literature review, following the PRISMA guidelines, is carried out, focusing on the current hypotheses in opportunistic maintenance, including component interactions, workers’ skills and resource constraints, economic dependence and optimization objectives. The different dependence types are identified and defined, and their presence in the literature is quantified. The included papers in this review (n=91) were selected on the basis of relevance to the research questions from the Web of Science, Scopus and Google Scholar databases. Exclusion criteria were set, related to the year of publication (from 2000) and language (limited to French or English), and inclusion criteria required the paper to cover modeling, simulating or reviewing literature related to opportunistic maintenance with dependencies. The results show that economic dependence is mostly modeled by sharing downtime or set-up costs. The objective function for optimization is mostly found to be the economic cost of maintenance, with concerningly little consideration for environmental indicators. These results are finally discussed in light of advances in predictive analytics and current challenges in the sustainability of industrial processes. Further developments should consider including the social and environmental aspects of sustainability in the dependencies, but also look into the benefits that predictive analytics can bring to opportunistic maintenance. The variety of modeling assumptions and dependences presented in the literature does not always allow comparing the results of the models
Attitudes, self-efficacy and behavioral intentions to exercise with a socially assistive robot in individuals with schizophrenia: An exploratory study
No full textInternational audienceIndividuals with schizophrenia generally engage in significantly less physical activities in comparison with the general population, partly due to motivation deficits, but also structural barriers such as lack of available qualified health professionals. One solution is to use social robots to facilitate engagement in exercise. This research has a dual focus: first, comparing attitudes, self-efficacy, and behavioral intentions towards engaging in exercise with NAO between individuals with schizophrenia and healthy controls; second, assessing the impact of interpersonal coordination between the robot and human on intention to engage in exercise in both groups. In this study based on health Behaviour Change Theories, 22 participants with schizophrenia were matched with 24 nonclinical participants. Physical activities were conducted with NAO under two conditions: Synchronizing (NAO is adaptive to the participant’s rhythm) and Not-Synchronizing (non-adaptive). Participants completed questionnaires measuring attitudes, self-efficacy, and intention to engage in exercise with NAO after each condition. Results indicated that both groups held generally positive attitudes towards exercise with NAO. However, individuals with schizophrenia reported lower self-efficacy compared to control participants but similar behavioral intention to engage in exercise with NAO than control participants. Behavioral intention subscales were positively correlated with positive attitudes and self-efficacy. This study contributes to the increased efforts to develop new technologies for use in mental health care interventions for adults with severe mental disorders. Finally, we will discuss the potentials and limitations of using social robots to promote exercise in individuals with schizophrenia
Modeling the Dynamics of Approach and Avoidance Motivation in Achievement Context with an Agent-based Model
No full textInternational audienceThe shape of the corresponding attractor landscape is assumed to evolve over time through interactions among three necessary and sufficient social-cognitive variables, namely competence expectancies, expected benefit for the self, and threat for the self. These variables interact within and across personal, contextual, and situational levels. The present study aimed to (a) develop an Agent-Based Models (ABM; Smaldino, 2023) capable of simulating these interactions and therefore the dynamics of approach and avoidance motivation patterns, and (b) compare the outputs of this ABM with longitudinal data from three athletes and two PhD students pursuing an important mid-term (1–2 years) goal. Whittle's Maximum Likelihood Estimator (Roume, 2023) was used to detect 1/f power-law distributions—a temporal variability typical of complex dynamical phenomena—in the time series of both simulation and ecological data. Findings revealed 1/f distributions in both types of time series, thus supporting the relevance of the CDS paradigm in understanding approach and avoidance motivation. They also pave the way for future research testing intervention hypotheses on motivation using computer simulation
Clustering for AI Explainability to Replace Simulation Models
No full textInternational audienceAirbus Helicopters uses simulation models to compute performance metrics such as delivery dates, investment, and work-in-progress. There are many parameters and scenarios to test, which require a lot of time to run and result in long computation times. This paper explores the result of AI model that aims to replace simulation models. Synthetic data, generated through simulations and genetic algorithms, is used to train the AI model. Although the AI model provides faster predictions, it does not provide an explanation for its predictions. Clustering helps uncover patterns and interpret these predictions, making AI more transparent and applicable in industrial contexts
Convex Mixture Criterion Based Evidential Set-Valued Classification
No full textInternational audienceImperfect data have to be processed in a different way when they are involved in classification or regression tasks. For instance, in sensitive domains, e.g., medical field, predicting a subset of candidate classes when imperfect data are present is preferable to predicting a single class using point prediction methods which offer less guarantee. Of course, the corresponding classifier should be cautious, i.e., the predicted subset of candidate classes contains the true class, but also relevant or precise, i.e., its size is not too large. This paper focuses on the adaptation of the Strong Dominance criterion (SD) to a more flexible classifier by combining it to the pignistic criterion (PC) within the framework of belief functions. Indeed, SD is a good candidate for cautious classification tasks as a robust method but in some situations it predicts subsets that are too large. The proposed set-valued classifier, based on a convex mixture (CM) between SD and PC, allows the control of the granularity of the partial order that is returned by SD regarding the desire of the decision-maker. The outputs of PC, SD and CM classifiers are theoretically studied and compared. Experimental results on fashion mnist data show that the proposed classifier might lead to better performances for the measures that more reward precision